Tavily Search

AI-optimized web search using Tavily API via web_fetch. Designed for AI agents — returns clean, relevant content without HTML parsing.

Search

web_fetch({
  url: "https://api.tavily.com/search",
  method: "POST",
  headers: {"Content-Type":"application/json"},
  body: "{\"api_key\":\"YOUR_TAVILY_API_KEY\",\"query\":\"YOUR_QUERY\",\"max_results\":5}"
})

Replace YOUR_TAVILY_API_KEY with the value from config get env.TAVILY_API_KEY.

Parameters (in body JSON)

Parameter	Description
query	Search query (required)
max_results	Number of results, 1-20 (default: 5)
search_depth	"basic" (default, fast) or "advanced" (deeper, slower)
topic	"general" (default) or "news"
days	For news topic, limit to last N days
include_answer	true to get an AI-synthesized answer

Response

{
  "results": [
    {
      "title": "Page title",
      "url": "https://example.com",
      "content": "Relevant extracted content",
      "score": 0.95
    }
  ],
  "answer": "AI-synthesized answer (if include_answer: true)"
}

Extract content from URL

web_fetch({
  url: "https://api.tavily.com/extract",
  method: "POST",
  headers: {"Content-Type":"application/json"},
  body: "{\"api_key\":\"YOUR_TAVILY_API_KEY\",\"urls\":[\"https://example.com/article\"]}"
})

Setup

1. Get API key at https://tavily.com
2. Store it: config set env.TAVILY_API_KEY your-key-here

Notes

• Free tier: 1,000 searches/month
• Uses web_fetch — works on all channels (no exec restriction)
• AI-optimized: returns clean snippets, not raw HTML
• include_answer: true gives a synthesized answer with citations